To me it could totally make sense. It depends on the context, but I can very well see contexts where such a choice could make sense. For example, in line of principle it would make sense, for an RSA implementation, to accept to construct a type PublicKey only computing the product of two Prime's, and not two arbitrary numbers. And the Prime type would only be constructible by procedures that provably (perhaps with high probability) generate an actual prime number. It would be a totally sensible form of defensive programming. You don't want to screw up your key generation algorithm, so it makes sense to have your compiler help you to not construct keys from anything.
For the same reason, say, in an HTTP server I could store a request as a char* or std::string, but I would definitely create a class that ensures, upon construction, that the request is valid and legitimate. Code that processes the request would accept HTTPRequest, but not char*, so that unverified requests cannot even risk to cross the trust boundary.
For the same reason, say, in an HTTP server I could store a request as a char* or std::string, but I would definitely create a class that ensures, upon construction, that the request is valid and legitimate. Code that processes the request would accept HTTPRequest, but not char*, so that unverified requests cannot even risk to cross the trust boundary.